33 way shootout- T.I.M.

[Richard Dower]

http://benchmarkreviews.com/index.php?option=com_content&task=view&id=138&Itemid=1

That's Thermal Interface Material in case you wondered ;)

[SiGfever]

Very interesting, I will try the T-C Grease 0098 if they return my inquiry.

[cegras]

Report results too - best to aggregate all findings so we can be sure it was the actual winner.

[hecktic]

they should have also included Liquid Metal TIM

[HousERaT]

I didn't care for the review much. Their findings with AS5 compared to what many other reviewers have found against newer TIMs is suspect.

[[XC] gomeler]

Would have liked to see a longer duration test. Give each TIM a good week of burn-in time. With 3 to 5 identical test-beds this wouldn't take too long and it'd give us a better idea of how the compounds set in over a period of time. Nice tests though and I appreciate the lengthy testing time, I feel their pain :yepp:

[Speederlander]

No liquid metal pro?

[situman]

All within margin of error. MX2 still rules in my opinioin.

[TyphooN]

I just bought some T-C Grease 0098. I have been using Arctic Silver 5 and Ceramique for a very long time. Hopefully this new paste performs better for me. It is only $4 + shipping for 10g of it, pretty cheap so I am willing to give it a shot.

[cajer]

Time to whip out the thermal paste that came with my TRUE. It is also T-C Grease 0098.

[mcoffey]

lame testing methodology IMO,

useless unless the TIM is allowed to cure. I waste of time.

andyc

[JaD]

This test is completely worthless.

Comparing temperatures with a difference as small as 0.2°C with a +-2°C accuracy thermal sensor (like the ones used inside cpu cores) that can't also report decimal values is just nonsense.
Everyone knows how CoreTemp readings can vary from second to second on the same load. So how can you come to any conclusion if 30 seconds later your reading might easily be a degree higher? How can you tell "XX's is best" and "YY's is not worth mentioning" while reporting an opposite situation going from idle to full load (when temperature, by just changing interface material, should show a solid difference)?

Not considering you really should lap both surfaces not to come into unfair situations where higher viscosity TIMs will take advantage over more liquid ones because of a convex HIS, wich is likely for almost any HIS.
Also look how curved (http://www.pctuner.info/up/image.php?src=_200702/20070226202039_CIMG1583.JPG) S775 retention clips are and how their pressure curves a CPU - open (http://www.pctuner.info/up/image.php?src=_200702/20070226201920_CIMG1581.JPG) / closed (http://www.pctuner.info/up/image.php?src=_200702/20070226201950_CIMG1582.JPG) socket

The part about scientific method being a good reason why they didn't take the time needed for TIMs to settle is just silly.
You need to give every product the same amount of cure time for an accurate comparison - right - and it would take too much time for every paste to cure 200 hours - right again - but seriously, because you don't have 200x33 hrs you test with not a single thermal cycle? LOL.
That "upto 200hrs" suggested by Arctic Cooling is the time needed for best performance under normal use. If you enable C1E and cycle 5 minutes of full load/idle, time needed for the paste to settle will be drastically reduced. Still an issue cycling for hours? Fine, do it just 5 or 6 times; possibly not enough to get best performance from some specific products, but better than nothing if you are looking into a product specifically designed for long term usage. Worth an additional day of testing, for way more fair results.

It's always like that, hardware reviewers just don't have the necessary equipment and knowledge to accurately test small temperature differences (and that goes for heatsinks/waterblocks aswell).

Sorry for the bad English WOT.

[SKYMTL]

I have to say that testing thermal compound is not an exact science by far...much less so when reporting maximum temperatures as reported by Core Temp or any other temperature monitoring program for that matter. Heck, nearly all of these temperature monitoring programs jump around like nuts from second to second so getting a true maximum reading is next to impossible.

I have found the best way to calculate temperatures is to run a constant load on the processor for a set amount of time while logging the temperature at 1-3 second intervals. Then, instead of reporting the maximum temperature, use the AVERAGE temperature for the entire benchmark run. It becomes very interesting when you start averaging things out instead of always talking about maximums.

Something that is also VERY important when testing thermal compound is to report on at LEAST two seperate mounts. We all know that one mount of a heatsink can report a FAR different temperature from another. This will impact the overall performance of a thermal compound quite a bit.

And yes, it is also extremely important to let the compounds settle for their dictated curing times. HOWEVER, there are compounds on the market (MX-2, Freeze, etc.) that claim they don't need any curing time. Thus, every compound should be tested with its manufacturer-suggested curing time but if a product says it doesn't need a curing time then it should be tested as such.

[JoeBar]

No Zalman STG1 in the review... :down:

[RADCOM]

Here's an oldie from 2003 and last year. My trusted Ceramique was top notch then lol
http://www.overclockers.com/articles907/

http://www.madshrimps.be/?action=getarticle&number=5&artpage=3071&articID=635

[Das Capitolin]

This test is completely worthless.

Comparing temperatures with a difference as small as 0.2°C with a +-2°C accuracy thermal sensor (like the ones used inside cpu cores) that can't also report decimal values is just nonsense.
Everyone knows how CoreTemp readings can vary from second to second on the same load. So how can you come to any conclusion if 30 seconds later your reading might easily be a degree higher? How can you tell "XX's is best" and "YY's is not worth mentioning" while reporting an opposite situation going from idle to full load (when temperature, by just changing interface material, should show a solid difference)?

Not considering you really should lap both surfaces not to come into unfair situations where higher viscosity TIMs will take advantage over more liquid ones because of a convex HIS, wich is likely for almost any HIS.
Also look how curved (http://www.pctuner.info/up/image.php?src=_200702/20070226202039_CIMG1583.JPG) S775 retention clips are and how their pressure curves a CPU - open (http://www.pctuner.info/up/image.php?src=_200702/20070226201920_CIMG1581.JPG) / closed (http://www.pctuner.info/up/image.php?src=_200702/20070226201950_CIMG1582.JPG) socket

The part about scientific method being a good reason why they didn't take the time needed for TIMs to settle is just silly.
You need to give every product the same amount of cure time for an accurate comparison - right - and it would take too much time for every paste to cure 200 hours - right again - but seriously, because you don't have 200x33 hrs you test with not a single thermal cycle? LOL.
That "upto 200hrs" suggested by Arctic Cooling is the time needed for best performance under normal use. If you enable C1E and cycle 5 minutes of full load/idle, time needed for the paste to settle will be drastically reduced. Still an issue cycling for hours? Fine, do it just 5 or 6 times; possibly not enough to get best performance from some specific products, but better than nothing if you are looking into a product specifically designed for long term usage. Worth an additional day of testing, for way more fair results.

It's always like that, hardware reviewers just don't have the necessary equipment and knowledge to accurately test small temperature differences (and that goes for heatsinks/waterblocks aswell).

Sorry for the bad English WOT.

You should be apologizing for making such a baseless statement, because there wasn't a single reference to the "dream test" which you described. I think that to call someone's hard work totally worthless is harsh, and makes you seem bias. It would also mean that you are claiming that nothing of value was offered by this comparison of 33 different products, such as which ones are complete garbage.

[Spawne32]

Im gona stay out of this one, however i will say that i do agree with some of the statements listed above which is why i sort of backed out of my initial plans for a thermal paste roundup. Heatsinks were far easier to test by comparison.

[Ubermann]

anyone can buy a review at some odd site..its just money.

[Das Capitolin]

anyone can buy a review at some odd site..its just money.

Huh? I don't understand what you intend to mean by "buy a review"? Please explain.

[hollo]

the curing thing is a good point - pastes with no cure time have an instant advantage over those that have no cure time if you don't let any of the pastes cure.

they should follow-up the article by doing a full-cure test of the top 5-10, cos the top places (the ones people are interested in) would definitely shuffle a little.

[Das Capitolin]

they should follow-up the article by doing a full-cure test of the top 5-10, cos the top places (the ones people are interested in) would definitely shuffle a little.

That's what is indicated at the end of the article:

Because of the overwhelming response to this article and with the assistance of manufacturer feedback, Benchmark Reviews has planned to revisit many of the top products which made it into round two of this project with similar test results. In this future "part II" article, each test will include one-hour of constant thermal cycling to allow for a basic cure time. Although this procedure would have been impractical with the large test group used in this article, it would be worth the effort to conduct on a more refined test group.

[hollo]

oh true, i'd skimmed the last page, mostly looking at the numbers, sorry :P

do you think there would be any chance of doing a "Part III" followup to "Part II" - a week-long (168 hour), automated (if possible? or maybe just switched on in the morning and off at night) full-load/power-off cycling cure for the 5 round 3 TIMs?
and... dare i ask... repeated twice for each TIM

it'd take a long time, 10 weeks, but it wouldn't require a whole lot of attention, or 5 weeks if you used two rigs.

pretty please with a cherry on top... if you can spare the equipment... enthusiasts would book-mark the page, shower your site with hits, and the good karma would bring you many reincarnations of good luck :D :wierd: :party:

[SKYMTL]

That's what is indicated at the end of the article:

Unfortunately, that may bring you more critique from even more quarters. Certain thermal compounds (OCZ Ultra 5+, AS Ceramique and AS Silver) require MUCH more curing time than 1 hour in order for them to reach their maximum thermal conductivity. Users have actually reported pretty dramatic differences after 200 hours of AS5 curing and 25 hours (or thereabouts) of Ceramique curing time.

Adding an hour of constant thermal cycling will not accomplish much IMO. You are better off taking the time to test these top products with their actual curing times and retest any compounds from Round #1 which need curing times in order to get truer, more accurate results.

[Das Capitolin]

Would you mind linking to where it is these manufacturers recommend curing times? The more and more I think about it, the less I think a re-test would accomplish.

Since I tested everything equally, then there is a lot to work from on that basis. If certain TIM's require special attention, then I would like to see that stated somewhere on the package or product... WHICH THEY DO NOT.

[noXqzs]

I wouldn't use a normal pc to test the material, but a controlled heat source. With a good regulated psu, it is possible to precisely control the amount of power in any type of resistive element. This type of testbed would ensure that the only difference in tests would be the quality of the T.I.M. being tested. Something like this would be reproducible quite easily, allowing for multiple tests at one time.

[hollo]

Would you mind linking to where it is these manufacturers recommend curing times? The more and more I think about it, the less I think a re-test would accomplish.

a big problem is that a quick test doesn't replicate actual usage conditions, where a single TIM application may last for months or years.

and there'd be no reason to complain about the methodology except that we already know that TIMs improve over time from direct experience and from certain manufacturers stating that this is the case.

if you would be willing to do subsequently longer tests on smaller and smaller groups of TIMs i think it would be interesting in several ways, but mostly because the rankings would undoubtedly change. AS5, as an example of a round 3 TIM, would improve - if the AS5 delta-T improves by 2 degrees, but the TIM consultants and OCZ freeze do not, then how misleading are the results as they stand given there is a 1.1 degree spread for the top 5?

aside from that - which TIMs benefit from curing and by how much is valuable information for overclockers, and again we only really need to know that information for the few best TIMs. for many TIM applications we'll know in advance that the TIM application is a temporary thing, for a benchmark run or similar, so using a TIM that benefits from curing is pointless.

if you were able to follow-up with a 1 hour cycle for round 2, and a several-day long cycle for round 3, the article would be one of the best reviews done on this subject :)

[SKYMTL]

Would you mind linking to where it is these manufacturers recommend curing times? The more and more I think about it, the less I think a re-test would accomplish.

Since I tested everything equally, then there is a lot to work from on that basis. If certain TIM's require special attention, then I would like to see that stated somewhere on the package or product... WHICH THEY DO NOT.

They do state it in numerous cases.

AS5

LINK (http://www.arcticsilver.com/as5.htm)

Due to the unique shape and sizes of the particles in Arctic Silver 5's conductive matrix, it will take a up to 200 hours and several thermal cycles to achieve maximum particle to particle thermal conduction and for the heatsink to CPU interface to reach maximum conductivity.


Ceramique

LINK (http://www.arcticsilver.com/ceramique.htm)

Due to the unique shapes and sizes of the particles in Céramique, it will take a minimum of 25 hours and several thermal cycles to achieve maximum particle to particle thermal conduction and for the heatsink to CPU interface to reach maximum conductivity. (This period will be longer in a system without a fan on the heatsink.)


OCZ Ultra 5+

LINK (http://www.ocztechnology.com/displaypage.php?name=ultra5instructions)

Please note that regardless of the processor used, OCZ Ultra 5+ will take approximately 150-200 hours before its optimal conduction is realized. It is perfectly normal to see a reduction of 1-5 degrees Celsius over this time period depending on how...


I have found that actually contacting the manufacturer is th ebest source of information about these things since quite a few companies do NOT publish this info.

[Das Capitolin]

a big problem is that a quick test doesn't replicate actual usage conditions, where a single TIM application may last for months or years.

and there'd be no reason to complain about the methodology except that we already know that TIMs improve over time from direct experience and from certain manufacturers stating that this is the case.

if you would be willing to do subsequently longer tests on smaller and smaller groups of TIMs i think it would be interesting in several ways, but mostly because the rankings would undoubtedly change. AS5, as an example of a round 3 TIM, would improve - if the AS5 delta-T improves by 2 degrees, but the TIM consultants and OCZ freeze do not, then how misleading are the results as they stand given there is a 1.1 degree spread for the top 5?

aside from that - which TIMs benefit from curing and by how much is valuable information for overclockers, and again we only really need to know that information for the few best TIMs. for many TIM applications we'll know in advance that the TIM application is a temporary thing, for a benchmark run or similar, so using a TIM that benefits from curing is pointless.

if you were able to follow-up with a 1 hour cycle for round 2, and a several-day long cycle for round 3, the article would be one of the best reviews done on this subject :)

I am thankful posts like this which help give me real idea's to solve the problem of real-world testing of TIM's. I realize that there will always be a few brain-numb mouth-breathers who see fit to simply call someone's work worthless; but it's good to see that this forum is rife with them. :)

In the next article which will utilize a much hotter-running CPU with fractional measurements, I will utilize several thermal cycles to thin and cure the material. What hasn't been decided is how long this should be done. Obviously 200 hours is way too much for an article, but I think an hour of constant high-temp no-temp cycles could certain suffice.

Finally, of the three items you linked for manufacturer recommendations, two come from Arctic Silver and one I didn't use in the review.

[sjohnson]

I've never seen a TIM shoot-out use GC Electronics Type 44 in their testing. I've been using it for years, for me it's beat out AS3 and the original Shin-Etsu. As small a difference as exists between AS3 and AS5, I haven't bothered. Maybe when I use my Type 44 up I'll try again but $10 for enough to TIM a couple hundred CPU's it'll be a while.

http://www.directron.com/type44.html

[SKYMTL]

I am thankful posts like this which help give me real idea's to solve the problem of real-world testing of TIM's. I realize that there will always be a few brain-numb mouth-breathers who see fit to simply call someone's work worthless; but it's good to see that this forum is rife with them. :)

In the next article which will utilize a much hotter-running CPU with fractional measurements, I will utilize several thermal cycles to thin and cure the material. What hasn't been decided is how long this should be done. Obviously 200 hours is way too much for an article, but I think an hour of constant high-temp no-temp cycles could certain suffice.

Finally, of the three items you linked for manufacturer recommendations, two come from Arctic Silver and one I didn't use in the review.

What you could always do is what we do: have the computer run in the background curing the compound while you work on another review with your primary rig. It may take more time but I'm sure it would make a very interesting read.

Take my word for it. The crowd here at XS is VERY hard to please. ;)

[e30Birdy]

i agree with thier curing times and such are kinda bogus because as they stated some take longer and some need non, but trying out that OCZ Freeze is gonna be my next try, right now i either use AS5 and Mx-1 where i have to say i have found MX-1 to keep my sys a lil colder but it really isnt by much!

[xoqolatl]

Check this out instead:
http://www.xtremesystems.org/forums/showthread.php?t=182778

[strange|ife]

pretty nice write up imo. This T-C Grease 0098 looks interesting, and large quanity..will try to find some. good to see AS5 pretty much still is king overall. I've tried several paste over the years and always end up using as5 again. Mx2 was actually worse for me by a deg or 2. It all depends on mounting pressure, and technique when applying..in short most of these TIM's are all the same. Except for the really crappy OE grease

[K404]

It would have nice to see a performance per gram (or ml) per unit money spent... I didnt see that in the review?

Testing thermal compounds "properly" is an absolute minefield and will never please everyone.

[kiwi]

liquid metal beats AS5 by about 3-5C :)

But you cant clean it off properly even with acetone.

[trans am]

gotta give 'em credit. its a lot of work. :clap:
..then again, that time couldve been on spent overclocking.

[tastymannatees]

Why not test with a thermocouple cut into the IHS? That's what Intel recommends on any testing of consequence, a standard that can replicate results between tests and can be used to reference against the internal diode.

It's easy to file a cut in the IHS with toolmaker or jewelers files. I 've done it several times. Takes some of the speculation and guesswork out of the results with another data point.

[Savuti]

As someone has previously stated, somewhat surprised that they didn't include Coollaboratory's Liquid Metal in this round up. Applied properly, at the least, it would have probably finished in the top 3.

[Spawne32]

i agree with thier curing times and such are kinda bogus because as they stated some take longer and some need non, but trying out that OCZ Freeze is gonna be my next try, right now i either use AS5 and Mx-1 where i have to say i have found MX-1 to keep my sys a lil colder but it really isnt by much!

i sent some ocz freeze to nick for testing and it did well in our tests too, im gona be ordering up a larger tube of it for myself for my next build.

http://www.overclockerspulse.com/hardware-reviews/thermal-paste-comparison-noctua-nt-h1-vs-ocz-freeze/

[Lestat]

i have never seen a curing time increase or decrease temps.
N E V E R

temps are what they are within the first hour or so of installing the TIM.

nothing changes over a month or so.
regardless of what Arctic Silver would have you believe.

Das Capitolin - Arctic Silver does recommend on their "how to install" paperwork to leave the TIM on for a period of something like 200 horus ?? god its been a while since i seen that statement but they do recommend you leaving it on for a certain amount of time.

like it or not you will never see anything worth talking about after a period of a couple of hours.

best thing to do is just heat it up,, then shut it down over night so it hardens a bit then heat it up like normal.. thats the best your going to get.

and i GUARENTEE you that you will never get the same results from one application of TIM to the other.

plus there ARE variations when using the "size of a pencil eraser" method in the center of the cpu IHS, vs the long strip across the IHS.

you will see temps change there because they are two different application methods that allows the grease to be pushed out differently across the IHS.